Method of retaining a media stream without its private audio content

Abstract

A method is disclosed that enables the handling of audio streams for segments in the audio that might contain private information, in a way that is more straightforward than in some techniques in the prior art. The data-processing system of the illustrative embodiment receives a media stream that comprises an audio stream, possibly in addition to other types of media such as video. The audio stream comprises audio content, some of which can be private in nature. Once it receives the data, the data-processing system then analyzes the audio stream for private audio content by using one or more techniques that involve looking for private information as well as non-private information. As a result of the analysis, the data-processing system omits the private audio content from the resulting stream that contains the processed audio.

1. A method for processing audio content in an audio stream that includes private information, comprising:
 receiving the audio stream, at a data-processing system, the audio stream including a first speech waveform produced by a first party on a call, and a second speech waveform produced by a second party on the call; 
 determining whether a first segment of the first speech waveform comprises private information by analyzing the audio content of a segment of the second speech waveform against a model that describes the private information, wherein the model is represented by acoustic signatures in which non-prose-related characteristics of speech are stored; and 
 tagging the first segment when the first segment is determined to comprise private information. 
 
     
     
       2. The method of  claim 1  wherein the determining whether the first segment of the first speech waveform comprises private information is based at least in part upon a time location of the first segment within the first speech waveform. 
     
     
       3. The method of  claim 2  wherein the time location of the first segment is within the initial M seconds of the first speech waveform, M being a positive number that is less than the length in seconds of the first speech waveform. 
     
     
       4. The method of  claim 1  wherein the determining whether the first segment of the first speech waveform comprises private information is based at least in part upon a prosody of the segment of the second speech waveform. 
     
     
       5. The method of  claim 1  further comprising:
 determining whether at least one of a third segment and a fourth segment of the first speech waveform comprises non private information; and 
 retaining only the third segment where the third segment, but not the fourth segment, is determined to comprise non-private information. 
 
     
     
       6. The method of  claim 5  wherein the determining of the third segment is based at least in part upon a matching of the audio content of the third segment with at least one of a keyword and a key phrase. 
     
     
       7. The method of  claim 5  wherein the determining of the third segment is based at least in part upon a time location of the third segment within the first speech waveform. 
     
     
       8. The method of  claim 7  wherein the time location of the third segment is within the final N seconds of the first speech waveform, N being a positive number that is less than the length in seconds of the first speech waveform. 
     
     
       9. The method of  claim 1  further comprising altering the first segment when tagged. 
     
     
       10. The method of  claim 9  wherein the altering comprises substituting a tone in place of the first segment. 
     
     
       11. The method of  claim 9  wherein the altering comprises substituting silence in place of the first segment. 
     
     
       12. A method for processing private audio content in an audio stream, comprising:
 receiving the audio stream, at a data-processing system, the audio stream including a first speech waveform produced by a first party on a call, and a second speech waveform produced by a second party on the call; 
 determining whether a first segment or a second segment of the first speech waveform comprises non private information by analyzing the audio content of a segment of the second speech waveform against a model that describes private information, wherein the model is represented by acoustic signatures in which non-prose-related characteristics of speech are stored; and 
 retaining only the first segment when the first segment, but not the second segment, is determined to comprise non-private information. 
 
     
     
       13. The method of  claim 12  wherein the determining whether the first segment comprises non-private information is based at least in part upon a matching of the content of the first segment with one or more keywords or key phrases. 
     
     
       14. The method of  claim 13  wherein the determining whether the first segment comprises non-private information is also based at least in part upon a time location of the first segment within the first speech waveform. 
     
     
       15. The method of  claim 14  wherein the time location of the first segment is within the final N seconds of the first speech waveform, N being a positive number that is less than the length in seconds of the first speech waveform. 
     
     
       16. A method for processing private audio content in an audio stream, comprising:
 receiving the audio stream, at a data-processing system, the audio stream including a first speech waveform produced by a first party on a call, and a second speech waveform produced by a second party on the call; 
 creating a model describing the private audio content by generating an acoustic signature of the audio stream, in which at least one non-prose-related characteristic of speech is stored and in which intelligible prose has been omitted from the acoustic signature; and 
 retaining the acoustic signature as a representation of the audio stream. 
 
     
     
       17. The method of  claim 16  wherein the acoustic signature embodies at least one characteristic of the audio stream that is selected from the group consisting of call length, time location of one or more silent passages, and intonation. 
     
     
       18. The method of  claim 16  wherein generating an acoustic signature of the audio stream comprises low-pass—filtering the audio stream. 
     
     
       19. The method of  claim 16  wherein generating an acoustic signature of the audio stream comprises calculating pitch and energy of the audio stream. 
     
     
       20. The method of  claim 16  further comprising retaining a first segment of the first waveform in unaltered form, the first segment having been determined as comprising non-private information.